-
Notifications
You must be signed in to change notification settings - Fork 125
Add VDDK support in VMware to KVM migration #640
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: 4.22
Are you sure you want to change the base?
Changes from all commits
f8594f3
08e7427
dac9313
53ddd64
File filter
Filter by extension
Conversations
Jump to
Diff view
Diff view
There are no files selected for viewing
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -117,6 +117,179 @@ If you are hitting the following error when running ovftool, install the depende | |
| dnf install libnsl | ||
|
|
||
|
|
||
| VDDK-based Optimized Conversion | ||
| ------------------------------- | ||
|
|
||
| CloudStack supports an optimized VMware-to-KVM migration path using virt-v2v in vpx input mode combined with | ||
| VMware's Virtual Disk Development Kit (VDDK). This method eliminates the OVF export phase entirely and streams | ||
| disk blocks directly from the source hypervisor into the conversion pipeline, resulting in significantly faster | ||
| migration times. | ||
|
|
||
| The traditional OVF-based workflow operates in two sequential phases: | ||
|
|
||
| 1. Export the entire VM as OVF/VMDK files to temporary storage (full disk copy). | ||
| 2. Convert the local VMDK files using virt-v2v (second full disk read and write). | ||
|
|
||
| The VDDK-based workflow replaces both phases with a single streaming pipeline: | ||
|
|
||
| - virt-v2v connects directly to vCenter via ``vpx://`` | ||
| - Disk blocks are read on demand via VDDK (using nbdkit internally as the translation layer between the | ||
| VDDK API and virt-v2v's NBD block device interface) | ||
| - Conversion and disk transfer happen concurrently | ||
| - Only allocated blocks are transferred; zero-filled and sparse extents are skipped | ||
| - No intermediate OVF or VMDK files are created | ||
|
|
||
| This reduces disk I/O amplification, eliminates temporary staging storage, and shortens end-to-end migration time. | ||
|
|
||
| .. note:: | ||
|
|
||
| CloudStack does not distribute VDDK, operators must download it separately. | ||
| Along with the new VDDK-based conversion method the traditional OVF-based method remains supported for environments. | ||
| Operators can choose the conversion method on a per-migration basis in the UI import wizard. | ||
| Host Prerequisites for VDDK-based Conversion | ||
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | ||
|
|
||
| To use VDDK-based migration, operators must prepare each KVM host that will run the conversion: install the conversion | ||
| tools, install VDDK manually, configure libguestfs, and verify host connectivity to vCenter/ESXi. | ||
|
|
||
| Example: prepare one KVM conversion host | ||
|
|
||
| **Step 1: Install the conversion stack (RHEL / Rocky / Alma Linux)** | ||
|
|
||
| :: | ||
|
|
||
| dnf install -y epel-release | ||
| dnf config-manager --set-enabled crb | ||
| dnf install -y virt-v2v libguestfs-tools libguestfs-xfs qemu-img nbdkit | ||
|
|
||
| **Step 2: Configure libguestfs backend as** ``direct`` | ||
|
|
||
| VDDK does not work with the default sandbox backend. Configure ``direct``: | ||
|
|
||
| :: | ||
|
|
||
| cat <<EOF > /etc/profile.d/libguestfs.sh | ||
| export LIBGUESTFS_BACKEND=direct | ||
| EOF | ||
|
|
||
| source /etc/profile.d/libguestfs.sh | ||
|
|
||
| This can also be configured persistently with ``libguestfs.backend`` in ``/etc/cloudstack/agent/agent.properties`` (see `Agent Properties for VDDK-based Conversion`_ below). | ||
|
|
||
| **Step 3: Download and install VDDK** | ||
|
|
||
| Download the VDDK tarball and extract it to a known location on the KVM host, for example, ``/opt/vmware-vddk``. | ||
| The resulting directory must contain the VDDK libraries under a ``lib64`` subdirectory. | ||
|
|
||
| :: | ||
|
|
||
| mkdir -p /opt/vmware-vddk | ||
|
Contributor
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. is this default path for vddk (expected by nbdkit)?
Contributor
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. confirm on the path here |
||
| tar -xf VMware-vix-disklib-9*.tar.gz -C /opt/vmware-vddk | ||
|
|
||
| Expected layout after extraction:: | ||
|
|
||
| /opt/vmware-vddk/vmware-vix-disklib-distrib/ | ||
| lib64/ | ||
| include/ | ||
| bin64/ | ||
|
|
||
| **Step 4: Add EL9 compatibility symlink (when using VDDK 9)** | ||
|
|
||
| On EL9 distributions, virt-v2v may expect ``libvixDiskLib.so.8``. Create this compatibility symlink: | ||
|
|
||
| :: | ||
|
|
||
| cd /opt/vmware-vddk/vmware-vix-disklib-distrib/lib64 | ||
| ln -s libvixDiskLib.so.9 libvixDiskLib.so.8 | ||
|
|
||
| .. note:: This compatibility symlink is commonly required on RHEL 9, Rocky Linux 9, and Alma Linux 9. | ||
|
|
||
| **Step 5: Verify host setup** | ||
|
|
||
| :: | ||
|
|
||
| ls /opt/vmware-vddk/vmware-vix-disklib-distrib/lib64/libvixDiskLib.so.8 | ||
| virt-v2v --version | ||
| nbdkit --version | ||
|
|
||
| **Step 6: Verify required network access** | ||
|
|
||
| The KVM conversion host must be able to reach: | ||
|
|
||
| .. cssclass:: table-striped table-bordered table-hover | ||
|
|
||
| ============================== ====== ============================== | ||
| Target Port Purpose | ||
| ============================== ====== ============================== | ||
| vCenter 443 API / authentication | ||
| ESXi hosts 902 VDDK NFC disk transfer | ||
| ESXi hosts 443 VM metadata | ||
| ============================== ====== ============================== | ||
|
|
||
| Agent Properties for VDDK-based Conversion | ||
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | ||
|
|
||
| The following properties can be configured in ``/etc/cloudstack/agent/agent.properties`` on each KVM host to enable and tune the VDDK-based conversion. | ||
| After editing this file, restart the CloudStack agent (``systemctl restart cloudstack-agent``). | ||
| These values can also be passed in details parameters in importVm API as key-value pairs. | ||
|
|
||
| .. cssclass:: table-striped table-bordered table-hover | ||
|
|
||
| +------------------------+----------------------------------------------------------------------+-------------------------------------------------------+ | ||
| | Property | Description | Default / Example | | ||
| +========================+======================================================================+=======================================================+ | ||
| | ``libguestfs.backend`` | The libguestfs backend for VDDK-based conversion. | ``direct`` | | ||
| | | Must be set to ``direct`` for VDDK to work correctly. | | | ||
|
Contributor
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. if "direct" is must for VDDK to work, should we avoid this config?
Contributor
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. remove this property as per the updated code.
Contributor
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. @harikrishna-patnala can you check this, is this property still valid? |
||
| +------------------------+----------------------------------------------------------------------+-------------------------------------------------------+ | ||
| | ``vddk.lib.dir`` | Path to the VDDK library directory on the KVM host. | ``/opt/vmware-vddk/vmware-vix-disklib-distrib`` | | ||
| | | Passed to virt-v2v as ``-io vddk-libdir=<path>``. | | | ||
| +------------------------+----------------------------------------------------------------------+-------------------------------------------------------+ | ||
| | ``vddk.transports`` | Ordered VDDK transport preference. | Example: ``nbd:nbdssl`` | | ||
| | | Passed as ``-io vddk-transports=<value>`` to virt-v2v. | | | ||
| +------------------------+----------------------------------------------------------------------+-------------------------------------------------------+ | ||
| | ``vddk.thumbprint`` | Optional vCenter SHA1 thumbprint. | If unset, CloudStack computes it automatically on | | ||
| | | Passed as ``-io vddk-thumbprint=<value>`` to virt-v2v. | the KVM host via ``openssl``. | | ||
| +------------------------+----------------------------------------------------------------------+-------------------------------------------------------+ | ||
|
|
||
| Example configuration in ``/etc/cloudstack/agent/agent.properties``: | ||
|
|
||
| :: | ||
|
|
||
| # LIBGUESTFS backend to use for VMware to KVM conversion via VDDK (default: direct) | ||
| libguestfs.backend=direct | ||
|
|
||
| # Path to the VDDK library directory for VMware to KVM conversion via VDDK, | ||
| # passed to virt-v2v as -io vddk-libdir=<path> | ||
| vddk.lib.dir=/opt/vmware-vddk/vmware-vix-disklib-distrib | ||
|
|
||
| # Ordered VDDK transport preference for VMware to KVM conversion via VDDK, passed as | ||
| # -io vddk-transports=<value> to virt-v2v. Example: nbd:nbdssl | ||
| # vddk.transports=nbd:nbdssl | ||
|
|
||
| # Optional vCenter SHA1 thumbprint for VMware to KVM conversion via VDDK, passed as | ||
| # -io vddk-thumbprint=<value>. If unset, CloudStack computes it on the KVM host via openssl. | ||
| # vddk.thumbprint= | ||
|
|
||
|
|
||
| Recommendations for Using VDDK-based Conversion | ||
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | ||
|
|
||
| **Use a single primary storage pool for direct conversion** | ||
|
|
||
| When VDDK-based conversion is enabled, it is strongly recommended to configure the conversion to write directly | ||
| to the destination primary storage pool (i.e., set *Convert to storage pool directly* to ``true`` in the import wizard). | ||
| This eliminates the two-step process of the traditional OVF method, conversion to temporary storage followed by | ||
| an import step, replacing it with a single streaming pipeline that writes converted QCOW2 disks directly to the | ||
| destination primary storage. | ||
|
|
||
| **Network placement for optimal disk transfer throughput** | ||
|
|
||
| For best performance, place the KVM conversion host on the same high-bandwidth network as the source ESXi hosts. | ||
| VDDK disk transfer uses VMware's NFC protocol on TCP port 902. ESXi routes NFC traffic to the conversion host based | ||
| on standard IP routing, if the conversion host is reachable over a dedicated storage or migration network, | ||
| ESXi will naturally select that VMkernel interface for disk transfer, keeping bulk data off the management network | ||
| without requiring any special configuration in virt-v2v or CloudStack. | ||
|
|
||
| Usage | ||
| ----- | ||
|
|
||
|
|
||
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Step 2 may not be required if the property is not valid